Rosetta@home on Raspberry Pi and other ARM devices

Questions and Answers : Unix/Linux : Rosetta@home on Raspberry Pi and other ARM devices

To post messages, you must log in.

AuthorMessage
dcarrion87

Send message
Joined: 13 Oct 12
Posts: 1
Credit: 15,707
RAC: 0
Message 74474 - Posted: 19 Nov 2012, 8:29:58 UTC

Hello

Just poking to see if there is a compatible version of Rosetta application for ARM devices in the works? I already have SETI@home running optimized on my ARM devices thanks to their available source code. It\'s a shame that this project does not provide source code.

I\'m currently getting ~320 floating point MIPS (Whetstone) per CPU on a Raspberry Pi. These ARM devices are only going to get more popular and more powerful. I think it\'s worth engaging the community to test an ARM version if it\'s in the works.

Regards

Daniel

ID: 74474 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator
Project administrator

Send message
Joined: 22 Aug 06
Posts: 3520
Credit: 0
RAC: 0
Message 74495 - Posted: 20 Nov 2012, 15:32:11 UTC

No, these devices are too short on memory to effectively run an application like Rosetta@home.
Rosetta Moderator: Mod.Sense
ID: 74495 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Chugumoto

Send message
Joined: 7 Apr 07
Posts: 1
Credit: 506,836
RAC: 0
Message 74674 - Posted: 6 Dec 2012, 18:02:17 UTC

ok... RPi...
and that about 2 core Cortex A9 1.6 GHz with 1Gb RAM?
1008 Whetstone, 2464 Dhrystone on nativeboinc test
ID: 74674 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Brian

Send message
Joined: 9 Dec 12
Posts: 3
Credit: 154,931
RAC: 0
Message 74746 - Posted: 17 Dec 2012, 3:16:18 UTC

First, I would like to say that some programmers and users are waaayyyyy to used to haveing all of the RAM and other system hardware we have today. Someone for someone to say that a divice to tooo limited has never tried to use such a device. You have to program more efficiently and write smaller programs.

Second, if you broke up a work unit into smaller sizes and wrote code for doing just certain aspects, then a RPi and other ARM cores would work. Think small, not big. If I had 4 RPi\'s on my desk, then 2 could be doing one task while 2 others were doing a different task to crunch a work unit.

Here is someone that asks a question about utilizing a processor and you shoot him down. You ask for solutions in various places to be able to get more people/processors to crunch for you and you say NO! when someone offers a solution. You should embrace such an Idea no matter how silly it maybe to you, but to others that understand something maybe you do not it leaves us scratching our head!!!!!
ID: 74746 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator
Project administrator

Send message
Joined: 22 Aug 06
Posts: 3520
Credit: 0
RAC: 0
Message 74748 - Posted: 17 Dec 2012, 15:13:39 UTC

I\'m sorry you took it that way Brian. Yes, my reply was too brief. The topic has been discussed many times before, with various potential new hardware discussed. I should point out that anything I say is the opinion of a volunteer forum moderator, and not a BakerLab Team member. But I know that when you\'ve been developing a computer program for over 10 years, to perform advanced science and achieve results that noone else in the world is achieving, you sometimes are unable to \"think small\". The goal here is to produce accurate models, not to be able to run on any given new piece of hardware.

If you can attempt 3 possible ways of producing more accurate models using lots of memory, or only 1 possible way that is able to run on a smaller computing environment, you have to err on the side of trying more possibilities. And this is the challenge with such programming projects. The science team needs to be working on the science, not the runtime footprint.
Rosetta Moderator: Mod.Sense
ID: 74748 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Brian

Send message
Joined: 9 Dec 12
Posts: 3
Credit: 154,931
RAC: 0
Message 74758 - Posted: 20 Dec 2012, 0:44:13 UTC

R@H is getting ready to miss the boat. Adapteva has release a dual ARM core chip with one of their 16 core Epiphany chips for parallel programming. They are also releasing their 64 core version. Granted, these chips don\'t have a lot of memory yet, but they are working on it. Plans are for a 16,000 core chip. Just think of the possibilites to be able to parallel process a work unit.

As an example, lets use SETI. If my computer and their software can only look at 1 freq. at a time, many times, just very fast, think if you could analize 16 freq. at the same time very fast! Now, imagine processing 16 folds or energy usage calculations at the same time.

Oh, by the way, the 64 core chip uses less than 5W of power. Think how much a GPU uses or a computer in general. How many of us have at least a 500W power supply in our system. Now, at 5W for 64 cores, that would relate to 100 64 core chips with each chip processing on the order of about 100gflops. Now, do the math again and 100gf x 100 cores is 10K gflops. Oh, and all of these processors would fit well within the confines of a shoe box!

Now, can we have some possible talk about R@H\'s future please? with sugar on top! Ha Ha Ha

Sincerely,
Brian
ID: 74758 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Paul

Send message
Joined: 29 Oct 05
Posts: 190
Credit: 55,023,485
RAC: 75,619
Message 78048 - Posted: 19 Mar 2015, 13:13:37 UTC

Just to bump this thread.

Raspberry Pi 2 has a quad core processor. While it is still limited to 1GB of RAM, that limitation is likely to improve.

Is there any effort underway to support this tiny powerhouse of a platform?
Thx!

Paul

ID: 78048 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
fechter

Send message
Joined: 25 Sep 14
Posts: 1
Credit: 843,849
RAC: 0
Message 78106 - Posted: 11 Apr 2015, 13:08:45 UTC

I\'d also be interested in an answer to this question.
ID: 78106 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Jonathan Jeckell

Send message
Joined: 17 Dec 05
Posts: 7
Credit: 3,994,686
RAC: 424
Message 78876 - Posted: 1 Oct 2015, 20:21:18 UTC - in response to Message 74758.  

R@H is getting ready to miss the boat. Adapteva has release a dual ARM core chip with one of their 16 core Epiphany chips for parallel programming. They are also releasing their 64 core version. Granted, these chips don\'t have a lot of memory yet, but they are working on it. Plans are for a 16,000 core chip. Just think of the possibilites to be able to parallel process a work unit.

As an example, lets use SETI. If my computer and their software can only look at 1 freq. at a time, many times, just very fast, think if you could analize 16 freq. at the same time very fast! Now, imagine processing 16 folds or energy usage calculations at the same time.

Oh, by the way, the 64 core chip uses less than 5W of power. Think how much a GPU uses or a computer in general. How many of us have at least a 500W power supply in our system. Now, at 5W for 64 cores, that would relate to 100 64 core chips with each chip processing on the order of about 100gflops. Now, do the math again and 100gf x 100 cores is 10K gflops. Oh, and all of these processors would fit well within the confines of a shoe box!

Now, can we have some possible talk about R@H\'s future please? with sugar on top! Ha Ha Ha

Sincerely,
Brian


I\'m going to try to install Android on a Raspberry Pi 2 this weekend--people have done it--and try to install Rosetta from there. The problem is the Android builds for the Pi I\'ve seen lacked the Google Play store...so I don\'t know how I will install BOINC without that.

I saw a reference to obtaining the source code elsewhere in the thread. I\'m not sure that will do me any good unless the compiler can make the code executable for ARM (probably not if it\'s optimized for other platforms, right?). Someone made an ARM/Raspberry Pi friendly version of SETI that works nicely.

Either way, I am making non-trivial contributions to other BOINC projects with numerous Raspberry Pi 2s (and soon some Orange Pi PCs). I can get more CPU power from $35 Raspberry Pi 2s or $15 Orange Pi PCs than a $500 PC.

Sure would be nice if this was supported.

ID: 78876 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mixu

Send message
Joined: 16 Jan 16
Posts: 1
Credit: 2,810,662
RAC: 0
Message 79783 - Posted: 22 Mar 2016, 21:55:12 UTC
Last modified: 22 Mar 2016, 21:57:00 UTC

*Bump*

500$ could get 100x RPi Zeros each with 1GHz ARM 11 and 512MB of RAM.
They eat so little power, that the entire cluster could easily be solar-powered and thus duct-taped to the backside of a single full-scale solar panel.

Another thing I\'d like to point out is that the SOC also contains a gpu that\'s capable of rendering decent 3d as well as full-hd video.

Both the cpu and gpu of each RPi enabled for computation, at that scale, could actually reach satisfying results.

Considering the power bill caused by a full rack of Xeons (..and their combined acquisition price), it might be easier to get more devices involved with an ARM-compiled and optimized version of the project.

Given that an RPi cluster built for the purpose can be \"fire and forget\" --> hermetically sealed and thrown to the back yard, all the ARM-devices around the world might even provide a more solid backplane than the current plan of relying mostly on average-spec desktops and laptops when they\'re not in any other use.

The cumulative result of all the RPis could make a notable addition to the few top-resulting users running the project in their company data centers.
ID: 79783 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Bogdan

Send message
Joined: 20 Mar 17
Posts: 4
Credit: 493,242
RAC: 0
Message 81353 - Posted: 21 Mar 2017, 15:41:13 UTC

Can anyone tell me if RPi3 is working with this project? I have 5 devices ready to go but it looks like they\'re not downloading any tasks.

If someone has some tips to get this working I would be appreciated.

Thanks,
Bogdan
ID: 81353 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote

Questions and Answers : Unix/Linux : Rosetta@home on Raspberry Pi and other ARM devices



©2019 University of Washington
http://www.bakerlab.org